It always should have had the right business model where they paid for this access for AI training. They knew it was wrong but in their rush to be known they decided it was better to take without asking and then ask for forgiveness later. Regardless what happens now, people have already made a name for themselves swindling the likes of Microsoft out of it and will have long well-paying careers from it.

These models can still be trained on data that they're allowed to use, but I think that what we're seeing is that the better LLM services are probably trained with shocking amounts of private data, whereas the less performant probably don't use stolen data.

It certainly seems illegal, but if it is, then all search engines are too. They do the same thing. Search engines copy everything to their internal servers, index it, then sell access to that copyrighted data (via ads and other indirect revenue generators).

I'm glad you bring up Google Books in this. Those lawsuits in the early teens about this issue are really important. But two things bother me: Google really won the case, but then basically abandoned the project. It's still there, but a shell of what it used to be. I wonder if the case may be, even though they won, they really lost. Or it could be Google just abandoning another project because they never cared about it.

I think AI for searching books like Google books would be an a amazing use case, and really, it is t that much different than what Google books is: an index of all of the published words. In fact, I can imagine AI being able to help you figure out if this book has the info you actually need from the book. That's not what GPT is, but one could make one that could do it.

I am torn. I am sort of a GPT may sayer, but on the other hand, is it really all that philosophically different than what humans do? I don't think it is materially different, but it is a little.

The thing is these are two separate arguments.

One is whether or not training is infringement.

The other is whether or not there needs to be stricter filters on output to avoid copyright.

The second one is easy to both argue for and implement, it just means funneling money towards a fine tuned RAG model to detect infringement before spitting out content. I'd expect we'll be seeing that in the near future. It's similar to the arguments YouTube was doomed at its acquisition because of rampant copyright infringement, but they just created a tagging system and now people complain about over-zealous DCMA enforcement - generative AI will end up in the same place with the same results for cloud-based models.

The first is much more murky, and I'm extremely skeptical that the suits regarding it will be successful given the degree of transformation and the relative scope of the material in any given suit compared to the total training set. As well the purpose of the laws in the first place were to encourage creation, and setting back arguably the most powerful creative tool in history (particularly when it means likely being eclipsed by other nation states with different attitudes towards IP) doesn't seem all that encouraging.

If I were putting money on it, we'll see multiple rulings against training as infringement which will settle the issue, but we will see "copyright detection as a service" models pretty much everywhere for a short period until suddenly the use of generative AI by creatives is so widespread that its being unable to be copyrighted means business models shift from media as a product to a service.